multilingual processing

vdr-2b-multi-v1 is a multi-language embedding model launched by Hugging Face, specially designed for visual document retrieval. The model is able to encode document page screenshots into dense single vector representations, enabling search and querying of multilingual visually rich documents without the need for OCR or data extraction processes. Developed based on MrLight/dse-qwen2-2b-mrl-v1, it uses self-built multi-language query-images to train the data set. It is an upgraded version of mcdse-2b-v1 with more powerful performance. The model supports Italian, Spanish, English, French and German, has an open source multi-lingual synthetic training dataset of 500,000 high-quality samples, features low VRAM and fast inference, and performs well in cross-language retrieval.